Brain
© GettyThe crystals act like neurons in the brain
Researchers have developed microscopic crystals that act like artificial neurons in the best attempt yet at mimicking the workings of the human brain.

The discovery, published in the scientific journal Nature Nanotechnology, has been heralded as a breakthrough in the nascent science of quantum computing, in which data processing tasks are carried out by collections of "machines" little larger than an atom.

When clumped together into networks, it is thought the crystals will be able to work together to carry out tasks far beyond the capability of the current breed of computers. The new systems will be able to carry out hyper-advanced analysis of weather systems and markets, as well as completing tasks innate to human beings, such as instantaneous facial recognition.

Chaotic data

The team of scientists, from the computer giant IBM, made the breakthrough by placing a microscopic quantity of an "electropositive" chemical compound between two electrodes. The compound, a chalcogenide, has a special quality whereby its structure changes from non-conductive to conductive when a current is passed through it. When enough energy is received, the crystal emits an electrical signal, closely mimicking natural neural processes.

Individually, these chalcogenide crystals - their diameter a fraction the size of a droplet of water in a cloud of fog - would be quickly overwhelmed. But the team behind the discovery found that when clumped into large groups, these artificial cells could work together to process large amounts of chaotic data.

In this way, they replicate the workings of neurons in our brains, which link up to form networks that can process extensive information and carry out complicated tasks very quickly.

Complex tasks

Having a large number of individual units working together is believed to be the most expedient way of tackling complex tasks, making the microscopic size of the crystals key to their success.

Scientists have long battled with the idea of creating artificial neurons, often to be overcome by the role of random trial-and-error - another aspect of successful data processing.

As well as being able to solve "big and complex" data problems, IBM researcher Tomas Tuma said the revolutionary systems come with the added advantage of needing only "a small power and energy budget".